This is a nice Bayes learning opportunity. It’s reasonable to infer that a female-looking username makes someone more likely to be female, maybe twice as likely (not much more than that; this is the internet and people give themselves weird usernames all the time, and actual women may avoid using female-looking usernames in male-dominated forums to avoid drawing attention to their gender). However, the base rate of transsexualism, even within a community as unusual as LW, is still incredibly low and requires a lot of evidence to overcome (e.g. someone telling you they’re transsexual).
On Less Wrong in particular, I would assign a high likelihood to various permutations of “gothgirl” being ironic, rather than sincere self expression of the user.
Do you really think 1/3rd of users named gothgirl* would be male?
That isn’t the relevant number. The likelihood ratio is P(named gothgirl | female) / P(named gothgirl | male), not P(female | named gothgirl) / P(male | named gothgirl).
(for what it’s worth, I didn’t reason using base rates, I just remember an early comment by gothgirl420666 saying he was male and only took that name for the lulz)
This is a nice Bayes learning opportunity. It’s reasonable to infer that a female-looking username makes someone more likely to be female, maybe twice as likely (not much more than that; this is the internet and people give themselves weird usernames all the time, and actual women may avoid using female-looking usernames in male-dominated forums to avoid drawing attention to their gender). However, the base rate of transsexualism, even within a community as unusual as LW, is still incredibly low and requires a lot of evidence to overcome (e.g. someone telling you they’re transsexual).
Do you really think 1/3rd of users named gothgirl* would be male? I’d guess something like 1-10%, compared with 1-3% transsexualism on LW: http://lesswrong.com/lw/fp5/2012_survey_results/
On Less Wrong in particular, I would assign a high likelihood to various permutations of “gothgirl” being ironic, rather than sincere self expression of the user.
Yeah, sure. This is the internet. (Acknowledged that the base rate of transsexualism on LW is higher than I had expected.)
That isn’t the relevant number. The likelihood ratio is P(named gothgirl | female) / P(named gothgirl | male), not P(female | named gothgirl) / P(male | named gothgirl).
(for what it’s worth, I didn’t reason using base rates, I just remember an early comment by gothgirl420666 saying he was male and only took that name for the lulz)
Oh hey, what’s up?